CrashNet: an encoder\u2013decoder architecture to predict crash test outcomes

Mohamed Karim Belaid,Maximilian Rabus,Ralf Krestel

doi:10.1007/s10618-021-00761-9

Abstract

Destructive car crash tests are an elaborate, time-consuming, and expensive necessity of the automotive development process. Today, finite element method (FEM) simulations are used to reduce costs by simulating car crashes computationally. We propose CrashNet, an encoder–decoder deep neural network architecture that reduces costs further and models specific outcomes of car crashes very accurately. We achieve this by formulating car crash events as time series prediction enriched with a set of scalar features. Traditional sequence-to-sequence models are usually composed of convolutional neural network (CNN) and CNN transpose layers. We propose to concatenate those with an MLP capable of learning how to inject the given scalars into the output time series. In addition, we replace the CNN transpose with 2D CNN transpose layers in order to force the model to process the hidden state of the set of scalars as one time series. The proposed CrashNet model can be trained efficiently and is able to process scalars and time series as input in order to infer the results of crash tests. CrashNet produces results faster and at a lower cost compared to destructive tests and FEM simulations. Moreover, it represents a novel approach in the car safety management domain.

Highlights

Multi-national car manufacturers are hampered, in their race towards globalization, by the law requirements of each country
We present CrashNet, a deep neural network that uses as input multiple scalar features and the car’s acceleration time series: a 1-dimensional time series representing the deceleration of the car during a crash event
A direct comparison between CrashNet and finite element method (FEM) simulations using mean square error (MSE) scores is not possible given the following reason: In general, the accuracy of FEM simulation results varies in the early stages of the development process and is constantly improved due to validation steps, e.g., after performing the first physical crash tests

Summary

Introduction

Multi-national car manufacturers are hampered, in their race towards globalization, by the law requirements of each country. For each new car model, a large class of crash tests with different scenarios has to be fulfilled These tests are performed multiple times during the development process to ensure that the required safety standards are being met. Crash cars are equipped with several hundred sensors, e.g., acceleration sensors on the car structure and inside the dummies The signals of these sensors can be understood as a time series and are used as the raw input data for CrashNet. But in contrast to large-scale perception tasks, these time series are accompanied by a set of scalar features like the car mass or the dummy type. In contrast to large-scale perception tasks, these time series are accompanied by a set of scalar features like the car mass or the dummy type They are used as additional information to predict crash test outcomes. CrashNet can be reshaped to output only one scalar such as e.g., the maximum chest acceleration

Objectives

Results

Discussion

Conclusion